In contemporary urban governance, the development of sustainable transportation requires a paradigm shift from reactive approaches to predictive ones. Bike-sharing systems, despite their environmental and social benefits, are persistently challenged by severe non-linear demand fluctuations that complicate optimal resource allocation. This study compares ten models—including time series (Prophet), machine learning (XGBoost, Gradient Boosting, Bagging), and deep learning (MLP, CNN, LSTM, GRU, Transformer)—for hourly demand forecasting in the Seoul bike-sharing system over horizons of 1, 3, 6, 12, and 24 hours. Empirical findings demonstrate that, based on the average performance across the evaluation metrics, the weighted tree ensemble model delivers the best and most stable results among all evaluated methods, achieving an R² of 0.962 at the one-hour horizon and 0.799 at the 24-hour horizon, while reducing prediction error by up to 54% compared with classical time-series methods. Policy implications for Iran's planning system include: (1) transitioning from linear to dynamic, prediction-based budgeting for clean transport projects; (2) establishing a national urban short-trip database; (3) adopting ensemble models as decision-support system cores; and (4) designing adaptive contracts with private operators. This approach provides an operational template for improving urban governance and reducing supply-demand imbalances in similar domestic schemes.The policy implications of this research for Iran’s planning system are: (1) the necessity of transitioning from linear budgeting to dynamic, forecast-based budgeting in clean transportation projects; (2) the importance of creating and standardizing hourly databases of weather variables and demand patterns in metropolitan areas; and (3) the feasibility of using hybrid models as the core of operational decision-support systems for managing bike-sharing fleets. This approach can provide an operational model for improving urban governance and reducing supply-demand imbalances in similar domestic initiatives (such as Tehran’s "Boodak" system).